Skip to main content
OpenSRE takes an alert or incident description, gathers evidence from your connected tools, and produces a root-cause report. Before you start: configure an LLM and set up the integrations you want checked (opensre onboard or opensre integrations setup).

​
Start an investigation

​
Interactive shell

Then either describe the incident in plain language, or run:
Bare /investigate opens a picker when you’re on a TTY. Templates include names like generic, datadog, and grafana. Full slash-command list: Shell commands.

​
CLI

Useful flags:

​
What you get

  • A report in the terminal (streams on a TTY when you’re not writing JSON)
  • Machine-readable JSON with -o ./rca.json (report, problem framing, root cause, and related fields)
  • In the shell after a run: /last to show it again, /save report.md or /save rca.json to write a file
  • A short summary to Slack, Telegram, or other chat tools if messaging is configured
If Slack is set up, OpenSRE can post a concise incident summary when the investigation finishes:
Slack Alert
Pipeline stages (intake → plan → gather evidence → diagnose → deliver): How an investigation works.

​
Background runs

In the shell, run /background on, then start an investigation. The run continues without blocking the prompt. This is per session only.